Post #1016: Probable cases revisited.

Posted on February 17, 2021

Source:  Calculated from Virginia Department of Health data, available at this URL.

This is a further analysis of a data reporting issue that I first came across in Post #1013. 

As it turns out, states vary in terms of what they count in their total of COVID cases.  In some states, totals include “probable” cases, mainly individuals who had a positive COVID antigen test, but not a PCR (DNA) test.  Other states do not include those “probable” cases found via antigen tests.

In particular, Virginia includes those cases, but adjacent Maryland does not.

This turns out to be a relatively minor issue for the U.S. as a whole.  The upshot of this is based on reporting by CNN:  The issue of under-reported COVID-19 positives, due to failure to report positive COVID antigen tests, is now limited to just five states:  CA, CO, MD, MO, NV.

And so, while the failure to report those tests is an annoyance, it probably has no material impact on measuring current national trends.  But this constitutes a significant caveat for the use of some official data sources.

Details follow.

In Post #1013, I stumbled across a data reporting issue when I noticed what a large fraction of new Virginia COVID cases were probable cases.  I was able to attribute that (mostly) to persons with positive antigen tests for COVID-19.   Those antigen tests are sometimes termed “rapid tests”, because the results are available much more quickly than PCR (DNA) tests.

(This reference is one of many plain-English explanations of the different types of COVID-19 tests.)

Those people — even though they had a positive antigen test — do not count as “confirmed” COVID-19 cases.  The only thing that counts as “confirmed” is a positive PCR (COVID-19 DNA) test.

And I noted that neighboring jurisdictions don’t include those probable cases in their counts.  That makes Virginia look like it’s doing materially worse than neighboring Maryland and DC.  When, in fact, we’re all doing about the same, in terms of new cases / 100,000 / day.

That was the first I’d heard of all that.  But now, it seems that I was just late to this party. 

First, this problem was pointed out months ago.  You can see writeups of it here, here, or here.  As the use of antigen tests has grown, that has (probably) resulted in an understatement of total new COVID-19 cases, relative to historical data.  In other words, an understatement of upward trend, and overstatement of downward trend.

Second, a reader pointed out that this split between confirmed and probable case is tracked by the U.S. CDC.  This information is available from this CDC dataset.  So, when I said I couldn’t tell which states are which, in terms of reporting positive antigen-based tests in their totals, to a first approximation, that was wrong.

I started to use that CDC file to get a handle on the extent of this problem.  But, given that this situation has been evolving, the exact state of the data isn’t crystal clear.  Looking in detail at the three largest states shows that and shows that there is no easy way to use the CDC at face value.

On the official CDC dataset, California, Texas, and Florida appear to have identical COVID case reporting.  The CDC file shows total cases but no breakout of confirmed (PCR test) versus probable (antigen test) cases.

That said, despite apparently identical reporting on the CDC file:

  • For California, the CDC total excludes antigen tests, because California does not report them.  California is one of five states (CA, CO, MD, MO, NV) that do not currently include antigen tests in official statistics, per this recent reporting from CNN.  (By inference, all the other states then include that in publicly-reported totals.)
  • For Texas, the CDC total includes antigen tests, even though Texas does not include them in its official total, by summing up the separately reported data from Texas.  Texas separately reports confirmed (PCR-test) cases and probable (antigen-test) casesThey began providing separate information on probable (antigen-test) cases as of December 11.  The CDC file reports the sum of confirmed and probable cases as reported by Texas, but not the separate figures.  (I note that the Johns Hopkins/NY Times data that I have been using also adds in those probable cases.)
  • For Florida, the CDC total includes antigen tests because the Florida reported total includes them.  The Florida total includes both PCR and antigen tests, which are then broken out separately (per their daily report, available on this web page).  As with Texas, the CDC dataset total is the combined count of PCR and antigen tests, with no separate breakout of “probable” cases.

I did not check the remainder of the file, but I would assume that all states not on the CNN list (plus, I think, DC) have antigen tests reflected in their totals.

And so, at the end of the day, this is an issue that affects California, and four other states, per the CNN reporting cited above.  Mostly, then, it’s a caveat to keep in mind if using the CDC data file.  And to bear in mind that, all along, what has been reported as the U.S. total has been a mix of states that did and did not include antigen tests in their COVID-19 case totals.